Introduction to recommender systems
0 likes748 views
Recommender systems are crucial for businesses and aim to predict user preferences to alleviate information overload by providing tailored recommendations. Various techniques such as content-based, collaborative filtering (both memory-based and model-based), and demographic or community-based approaches are utilized to enhance user experience. Despite their prevalence and importance, challenges persist in scalability, privacy, and diversity of recommendations.
Introduction to recommender systems
- 1. Introduction to Recommender Systems Rami Al-Salman
- 3. What is a recommender system? A recommender system, or a recommendation system is a subclass of information filtering system that seeks to predict the "rating" or "preference" a user would give to an item. (Wiki) The goal of a Recommender System is to generate meaningful recommendations to a collection of users for items or products that might interest them. (Melville, Sindhwani) Recommender systems reduce the information/choice overload by estimate the relevance
- 8. Business value of recommender systems Netflix: 75% of the movies watched Amazon: 35% sales Youtube : 38% more clickthrough https://arxiv.org/pdf/1908.08328.pdf
- 9. Sources of information for recommender systems Explicit ratings e.g, 5-stars, 1 star, etc Explicit binary ratings (thumbs up/thumbs down) Implicit information, e.g., how many times was it viewed? User profiles/preferences Item features
- 10. Recommender System general overview Recommendations Engine Id score 6 0.9 2 0.88 555 0.6 666 0.3 Recommendations Users Profiles & Contextual parameters Item Features
- 11. Classes of recommendation approaches (Burke, R, 2007) Content-based Collaborative filtering Demographic Community-based Hybrid approach
- 12. Content-based When a user liked an item, we try to recommend items that are similar to the the user liked in the past. The recommended items share some characteristics and features values with the liked item. If the user watched a horror movie and liked it, then next time we recommend horror movies with similar features i.e., description. Watched by a user Similar movies Recommended to user
- 13. Collaborative filtering (CF) Recommends to the active user the items that other users with similar tastes liked in the past. The similarity in taste of two users is calculated based on the similarity in the rating history of the users. CF is the most used and famous technique Watched by both users Similar users Watched by him, recommended to her
- 14. Collaborative Filtering (CF) Memory-based ● User-User ● Item-Item Model based: ● Matrix Factorization ● Deep learning
- 15. Memory-based (User-User) Harry Potter 1 Harry Potter 2 Host The Hunt User1 ? ? User2 ? User3 ? ? User4 ? ? ? Given users and movies, will user User1 like the Host and The Hunt movies?
- 16. Memory-based (User-User) Harry Potter 1 Harry Potter 2 Host The Hunt User1 1 1 ? ? User2 1 1 1 ? User3 ? 1 1 ? User4 ? ? ? 1 Get the most similar N-neighbour users. For simplicity, lets N to be 1 Given the most similar user, use what he/she liked to recommend to the source user For similarity, several functions can be used, e.g., Jaccard Similarity, Cosine similarity, etc
- 17. Memory-based (User-User) Harry Potter 1 Harry Potter 2 Host The Hunt User1 1 1 ? User2 1 1 1 ? User3 ? 1 1 ? User4 ? ? ? 1 Jaccard Similarity: Sim(A,B) = |A ∩ B| / |A| + |B| - |A ∩ B| Sim(User1, User2) = 2 / 3 Sim(User1, User3) = 1 / 3 Sim(User1, User4) = 0 Based on that we can recommend Host movie from User2 to User1 1
- 19. Demographic We recommend items based on the demographic profile of the user. The assumption is that different recommendations should be generated for different demographic niches. E.g., adapte language or country for specific e-commerce website
- 20. Community-based This type of system recommends items based on the preferences of the users friends. This technique follows the epigram “Tell me who your friends are, and I will tell you who you are”. (Bellotti, V, etl, 2008).
- 21. Challenges Scalability of the algorithms with large and real-world datasets Proactive recommender systems, i.e., recommenders that decide to provide recommendations even if not explicitly requested Privacy preserving recommender systems Diversity of the items recommended to a target user
- 22. Summary Recommender Systems are everywhere and critical for businesses to success Several kinds of recommender systems were presented Collaborative Filtering is most widely used technique There are still many open challenges to be tackled in the context of the recommender systems